1. Technical Field
The present invention relates in general to data processing systems and in particular to resource allocation within a data processing system. Still more particularly, the present invention relates to energy based resource allocation within a virtualized data processing system.
2. Description of the Related Art
In today's computer systems and data centers, specification of power consumption limits is becoming increasingly important. Specifying maximum energy consumption limits for a device is particularly important for several reasons: (1) There may be a maximum power supply to the building which houses a computer system executing critical applications; (2) the cooling capability of a data centre may only be sufficient to deal with a limited load; and (3) there may be a fixed budget for energy consumption based on financial cost, carbon emissions and other consumption based factors. If a consumption limit is to be employed on a single physical system in which multiple virtual machines/LPARs share resources, it is difficult to determine how to manage resources to keep power consumption within a specified consumption limit. Attempting to manage resources across multiple machines within a data center provides further difficulties.
Two existing approaches are the following: (1) A system is run at full power and is shut down immediately after the limit is reached; and (2) A system limits total energy usage and scales down the energy consumption gradually as the limit is approached. These existing approaches may result in energy resources being used by lower priority workloads at the expense of high priority workloads.
None of the existing alternatives allow any intelligent management of resources in light of an energy consumption limit. Intelligent management of resources may be especially important when multiple parties have a stake in the system. A data processing system that supports a cloud environment is a particular example of a system in which several different parties, groups or companies/stakeholders share physical resources.